AI-enhanced LF-MRI in Epilepsy (ALF-ME)

Evaluation of Artificial Intelligence-enhanced Low Field

The global burden of epilepsy is high affecting over 50 million people worldwide. Majority live in low- and middleincome countries (LMICs) where access to diagnosis and treatment is limited. Accurate diagnosis of epilepsy and identification of the underlying cause through brain imaging is key to providing appropriate treatment. Magnetic resonance imaging (MRI) is the recommended modality of choice for brain imaging. However, in many LMICs it is scarce, and the cost of maintenance is unattainable. This study aims to explore the usefulness of a lower cost, more portable MRI machine for epilepsy diagnosis. It will be a proof-of-concept study evaluating the utility of low magnetic field MRI (LF-MRI) in epilepsy diagnosis. It will include 30 adults with epilepsy who have undergone a high field MRI (HF-MRI) brain scan as part of their routine clinical care under the University College London (UCL) Hospitals (UCLH), within twelve months of recruitment. Participants will be consecutively recruited and offered a LF-MRI brain scan on the Swoop MR Imaging System (Hyperfine) at the Birbeck-UCL Centre for Neuroimaging (BUCNI). Image post-processing will be performed using the open access machine learning program, LF-SynthSR, to enhance the image quality and allow for quantitative image analysis. Anonymised HF- and LF-MRI scans will be independently reported using a structured reporting template by two neuroradiologists. A perception survey will be administered to all participants to assess their tolerability of the LF-MRI. This study will serve as a foundation for future studies in this field and in areas where such innovations are most needed.

Study Overview

• Status: Active, not recruiting
• Conditions: Epilepsy
• Intervention / Treatment: Diagnostic test: MRI

Detailed Description

The aim of this study is to determine the utility of AI-enhanced Low Field (AI-LF) MRI in the identification of focal brain lesions in adults with epilepsy. This is a proof-of-concept study evaluating the utility of AI-LF-MRI in aiding identification of potentially epileptogenic brain lesions, by comparing neuroradiologists' lesion detection on this modality versus standard HF-MRI. It will include ~30 adults with epilepsy who have undergone a high field (1.5T or 3T) MRI (HF-MRI) brain scan as part of their routine clinical care at University College London (UCL) Hospitals (UCLH), within the previous twelve months. Participants will be consecutively recruited and offered a LF-MRI brain scan on the 0.064T Swoop MR Imaging System (Hyperfine®) at the Birkbeck-UCL Centre for Neuroimaging (BUCNI). Image post-processing will be performed using the open access machine learning program, LF-SynthSR, to enhance the image quality and allow for quantitative image analysis. Anonymised HF- and LF-MRI scans will be independently reported using a structured reporting template by two neuroradiologists. A perception survey will be administered to all participants to assess their tolerability of the LF-MRI. Each participant will only have a single encounter with the study on the day they get their LF-MRI scan. There is no follow-up required with this study.

Inclusion criteria:

1. Adults aged between >18 years and <70 years
2. Diagnosis of epilepsy and attending a UCLH-affiliated outpatient epilepsy clinic
3. Undergone a 1.5T or 3T MRI Brain scan within 12 months of recruitment as part of standard clinical care.
4. Can tolerate MRI scanning without sedation. Exclusion criteria

1. Cognitive impairment that precludes ability to consent or assent to the study 2. Inability to lie flat for the duration of the scan 3. Body habitus incompatible with LF-MRI device 4. Presence of MRI contraindications as stipulated by standard MRI operating procedures

We will recruit at least 30 patients, of these, 20 patients will have a visible lesion on the HF-MRI scan (reference standard). A sample size of 20 patients will be sufficient to demonstrate that the proportion of lesions similarly identified is at least 0.8, against a null hypothesis of 0.5, using a one-sample, one-sided exact binomial test with a 5% significance level and a power of 80%. The rest will not have a visible potentially epileptogenic lesion on the HF-MRI scan.

Two consultant neuroradiologists will separately and independently report AI-LF- then HF-MRI scans according to set criteria, including, MRI scanner field strength, presence of lesion, lesion location and artefacts. They will be provided with basic participant clinical details such as age, seizure type and. EEG findings, but blinded to diagnosis. Rates of lesion detection and false positives in the AI-LF-MRI scans will be compared with the reference standard of a recent clinical HF-MRI and reported as proportions. Sensitivity and specificity analyses will be performed for the qualitative analysis of AI-LF-MRI. Intraclass coefficients and Bland Altman analyses will be used to compare quantitative inter-rater and inter-method agreement, respectively. Descriptive statistics such as medians (IQR) and proportions will be used to describe participants' demographic information and survey responses.

• Study Type: Interventional
• Enrollment (Actual): 39
• Phase: Not Applicable

Contacts and Locations

Study Locations

• United Kingdom
  • London, United Kingdom
    • UCL Queen Square Institute of Neurology

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

1. Adults aged between >17 years and <70 years
2. Diagnosis of epilepsy and attending at UCLH-affiliated outpatient epilepsy clinic
3. Undergone a 1.5T or 3T MRI Brain scan within 12 months of recruitment as part of standard clinical care
4. Can tolerate MRI scanning without sedation.

Exclusion Criteria:

1. Cognitive impairment that precludes ability to consent or assent to the study
2. Inability to lie flat
3. Body habitus incompatible with LF-MRI device
4. Presence of MRI contraindications as stipulated by standard MRI operating procedures

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Diagnostic
• Allocation: N/A
• Interventional Model: Single Group Assignment
• Masking: None (Open Label)

Number of Arms

1

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Experimental: Low-field MRI Scan; MRI Brain scan on the Hyperfine 0.064T Swoop System	Diagnostic test: MRI; Low magnetic field MRI of the brain; Other Names: Low-field

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Proportion of similarly detected lesions on AI-LF-MRI compared to HF-MRI.	The rate of correctly identified abnormalities on the AI-enhanced low-field MRI when compared to standard of care high-field MRI. There is no pre-specified "good" rate assigned for this pilot study.	Baseline assessment (within the study imaging visit). Participant will only undergo a low-field MRI scan which will be compared to a reference high-field MRI scan done within 12 months of enrolment.

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Proportion of artefacts on AI-LF-MRI compared to HF-MRI	Rate of reported artefacts on AI-LF-MRI compared to HF-MRI	Baseline assessment (within the study imaging visit)
Inter-rater agreement on qualitative analysis of AI-LF-MRI vs HF-MRI	Compare agreement of reports among radiologists reading the scans	Baseline assessment (within the study imaging visit)
Inter-method agreement on quantitative analysis of AI-LF-MRI vs HF-MRI	Comparative analysis of volume segmentation of the SuperSynth (AI tool) of low-field vs high-field MRI	Baseline assessment (within the study imaging visit)
Acceptability of LF-MRI scanning process	Describe the participant perceptions of their experience in the Hyperfine MRI scanner	Baseline assessment (within the study imaging visit)

Collaborators and Investigators

• Sponsor: University College, London
• Collaborators: University College London Hospitals; The Leverhulme Trust
• Investigators: Principal Investigator: John S Duncan, DM, University College, London

Publications and helpful links

Helpful Links

• SynthSR: A public AI tool to turn heterogeneous clinical brain scans into high-resolution T1-weighted images for 3D morphometry

Study record dates

Study Major Dates

• Study Start (Actual): September 18, 2025
• Primary Completion (Actual): February 20, 2026
• Study Completion (Estimated): August 31, 2026

Study Registration Dates

• First Submitted: July 7, 2026
• First Submitted That Met QC Criteria: July 15, 2026
• First Posted (Actual): July 21, 2026

Study Record Updates

• Last Update Posted (Actual): July 21, 2026
• Last Update Submitted That Met QC Criteria: July 15, 2026
• Last Verified: July 1, 2026

More Information

Terms related to this study

• Keywords: Epilepsy; Neuroimaging; Low- and middle-income countries; Low-field MRI; Resource-limited
• Additional Relevant MeSH Terms: Brain Diseases; Central Nervous System Diseases; Nervous System Diseases; Epilepsy
• Other Study ID Numbers: 352304; 179646 (Other Grant/Funding Number: University College London); 25/NE/0149 (Other Identifier: NHS Health Research Authority Reference number)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: UNDECIDED
• IPD Plan Description: Anonymised IPD may be shared with researchers within our institution (UCL). However, a plan has not yet been made for sharing the data more widely.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: Yes
• product manufactured in and exported from the U.S.: Yes